Can't Salesforce or HubSpot already detect at-risk deals?

Yes. Einstein and Breeze flag at-risk deals from activity data. What they don't do is model the specific human, diagnose why this champion pulled back, and prescribe the move. That's what Stage 2 adds with the GTM Heroes MCP.

What do I need to build this?

Stage 1 needs your CRM and Gmail connected in Claude Cowork. Stage 2 adds the GTM Heroes MCP, available on a free trial. Both run read-only on your CRM.

Will the agent change my deals?

No. Both prompts are read-only. The agent reads your pipeline and recommends. It never edits, writes, or moves a deal.

MCP Mondays Ep 1: Deal-Drift Detection

The setup

A committed-forecast deal. The demo crushed, the champion was fired up, next steps were booked. Then the replies slowed, the language went passive, and the economic buyer went dark three weeks ago. Your CRM still shows it green.

Stage 1 · your own tools

Flags the deal

Connect your CRM and Gmail. Build an agent that scores deal health off activity and surfaces stalls. Useful, and basically what Einstein and Breeze already do. It tells you a deal is slipping.

Stage 2 · + GTM Heroes MCP

Diagnoses the human

Add the GTM Heroes MCP. Now it reads the champion's Relationship Lens, names why they pulled back, scores the drift, and hands you the exact move to re-engage. Activity data can't do this.

Build it: Stage 1 (the OK version)

Needs: your CRM (HubSpot or Salesforce) + Gmail connected in Cowork. Read-only.

You are my deal-health watcher. Using my connected CRM and Gmail, read-only: 1. Pull my open deals in commit and best-case. 2. For each deal, compute: - days since the last inbound reply from the buyer - the trend in reply latency across the last 5 exchanges (speeding up or slowing down) - days since the last meeting - days in the current stage vs my average for that stage 3. Flag a deal as drifting if ANY of these are true: - reply latency is increasing - no inbound reply in 10+ days - stage age is more than 1.5x my average 4. Output a ranked watchlist: deal name, value, close date, the signals that tripped, and a one-line "why this looks stalled." Do not modify any deal. Read and report only.

Result: a clean stalled-deal watchlist. It tells you what is slipping. That's the ceiling of activity data.

Build it: Stage 2 (the way-better version)

Adds: the GTM Heroes MCP. Start a trial, grab MCP access, connect it in Cowork. Read-only on your CRM.

You are my deal-drift diagnostician. Using my CRM, Gmail, and the GTM Heroes MCP, read-only on my CRM: 1. Take the top stalled deal from my watchlist (or the deal I name). Call list_deals to confirm its stage, value, close date, and company. 2. Identify the champion on the deal. Call list_individuals to check if they are already profiled. If not, call profile_individual with their LinkedIn URL to get their HBX Relationship Lens and communication style. 3. Call research_prospect on the company for account context, persona prioritization, and the objection pre-butts. 4. Pull our engagement from my CRM activity + the Gmail thread. Read two things: the reply-latency trend, and the tonal shift in the champion's language (commitment language to hedging language). 5. Diagnose the disconnect. Compare how I have been selling this deal against the champion's Lens. Name the specific mismatch, then judge how far our recent engagement has drifted from the pattern that Lens rewards. 6. Prescribe the move. If there is a real disconnect, call generate_case_study for this individual (pick the most relevant proof from my Customer Proof Library) or generate_roi_calculator if the gap is economic, to produce a proof asset tuned to their Lens. Then tell me exactly how to frame the re-engagement. 7. Output: the named champion and their Lens, the disconnect in one sentence, the drift read, the prescribed next move, and a link to the generated asset. Do not modify any deal. Read, diagnose, and recommend only.

Result: "Your champion reads as The Architect... analytical, proof-driven. You led with vision. That's the mismatch, and why he went quiet. Here's a case study rewritten for The Architect, and how to frame the send."

Download the Deal-Drift skill (SKILL.md)

The delta is the point

Same deal, same data, two minutes apart. Stage 1 says the deal is at risk. Stage 2 says why this human is at risk and what to do about it. Run them back to back on one of your own stalled deals and the gap explains itself.

Run Stage 2 on your own deal

Start a trial, connect the MCP, point it at a deal that went quiet.

Start free & connect the MCP

The series

Live now

Ep 1 · Deal-Drift Detection

Next Monday

Ep 2 · Behavioral Profiling

Upcoming

Ep 3 · Lead Routing

Upcoming

Ep 4 · Playbook Generation

Get every Monday's build

New capability every Monday, 1pm ET. Register for the live series and we'll send each week's recipe.

About GTM Heroes.ai ... AI made outreach robotic. We use behavioral intelligence to make sales human again.